Merge "Bug 2055: Handle shard not initialized resiliently"

[controller.git] / opendaylight / md-sal / sal-distributed-datastore / src / main / java / org / opendaylight / controller / cluster / datastore / Shard.java

diff --git a/opendaylight/md-sal/sal-distributed-datastore/src/main/java/org/opendaylight/controller/cluster/datastore/Shard.java b/opendaylight/md-sal/sal-distributed-datastore/src/main/java/org/opendaylight/controller/cluster/datastore/Shard.java
index d0bb3d3b69824d18b2acdb08a4defb50d31953b2..d831b7c1844006fca5bea575ee20d656c4b616c4 100644 (file)
--- a/opendaylight/md-sal/sal-distributed-datastore/src/main/java/org/opendaylight/controller/cluster/datastore/Shard.java
+++ b/opendaylight/md-sal/sal-distributed-datastore/src/main/java/org/opendaylight/controller/cluster/datastore/Shard.java
@@ -217,6 +217,10 @@ public class Shard extends RaftActor {
 
         if (message instanceof RecoveryFailure){
             LOG.error(((RecoveryFailure) message).cause(), "Recovery failed because of this cause");
+
+            // Even though recovery failed, we still need to finish our recovery, eg send the
+            // ActorInitialized message and start the txCommitTimeoutCheckSchedule.
+            onRecoveryComplete();
         } else {
             super.onReceiveRecover(message);
         }
@@ -699,12 +703,15 @@ public class Shard extends RaftActor {
         //notify shard manager
         getContext().parent().tell(new ActorInitialized(), getSelf());
 
-        // Schedule a message to be periodically sent to check if the current in-progress
-        // transaction should be expired and aborted.
-        FiniteDuration period = Duration.create(transactionCommitTimeout / 3, TimeUnit.MILLISECONDS);
-        txCommitTimeoutCheckSchedule = getContext().system().scheduler().schedule(
-                period, period, getSelf(),
-                TX_COMMIT_TIMEOUT_CHECK_MESSAGE, getContext().dispatcher(), ActorRef.noSender());
+        // Being paranoid here - this method should only be called once but just in case...
+        if(txCommitTimeoutCheckSchedule == null) {
+            // Schedule a message to be periodically sent to check if the current in-progress
+            // transaction should be expired and aborted.
+            FiniteDuration period = Duration.create(transactionCommitTimeout / 3, TimeUnit.MILLISECONDS);
+            txCommitTimeoutCheckSchedule = getContext().system().scheduler().schedule(
+                    period, period, getSelf(),
+                    TX_COMMIT_TIMEOUT_CHECK_MESSAGE, getContext().dispatcher(), ActorRef.noSender());
+        }
     }
 
     @Override